Draft (July 10, 1995)
INTRODUCTION
The Foreign Exchange Committee ("the Committee") supports the effort of the Basle Committee to establish international guidelines for applying capital charges to the market risks incurred by banks. The Foreign Exchange Committee is encouraged that a number of its recommended changes to the April 1993 Basle Committee proposal on market risk ("BIS proposal") were incorporated in the current BIS proposal, most notably the use of banks' internal models for calculating market risk and the extension of Tier 3 capital to cover foreign exchange as well as other market exposures.However, the latest proposal, particularly the quantitative standards for using internal models, raises two major concerns:
 The extremely conservative quantitative standards in the proposal (including the multiplication factor, the holding period, restricted correlations and the confidence interval) would require banks to hold capital against daily price movements of as much as 24 standard deviations. Such capital requirements are out of proportion to actual risks in the foreign exchange market. The simple aggregation of capital for credit and market risks also overestimates the capital necessary for a diversified firm because potential losses from these risks are unlikely to be realized simultaneously. The additional costs imposed by such capital standards may shift a significant volume of trading activity to less regulated organizations.
 A regulatory model with minimum quantitative standards may actually impede progress toward developing more precise risk measurement systems. Virtually any comprehensive set of proposed quantitative standards will be in conflict with model parameters used by banks. Banks will continue to rely upon their own more precise internal models. The regulatory model, purely duplicative, will not be used in daytoday risk management and may prove impossible to validate using the proposed risk parameters. The regulatory model may divert resources from improvements to a bank's daytoday risk systems. The Committee strongly recommends that banks should be able to use their internal models as the basis for calculating regulatory capital requirements. Based upon their reviews of banks' internal models, regulators may adjust the model results, if necessary, using a multiplication factor greater than one.
QUANTITATIVE STANDARDS FOR USING INTERNAL MODELS
Members of the Foreign Exchange Committee understand the necessity of establishing conservative capital standards that capture a wide range of possible price movements. However, the BIS proposal assumes that each bank's portfolio is comprised entirely of the most illiquid and volatile traded instruments. In contrast, internal bank models are designed to more correctly reflect the actual composition of each bank's portfolio. The table below compares the current quantitative standards from the BIS proposal with parameters generally used by financial institutions. The cumulative effect of the BIS proposal standards is a total compounding factor ranging from 12.1 to 14.7, which is equivalent to a market move of approximately 24 standard deviations of daily price changes.1 Based on historical market volatilities in foreign exchange, Committee members believe that planning for at least 24 standard deviation price changes is unduly extreme.
If banks are required to maintain capital against "worst case" price movements while competitors' capital requirements are significantly lower, then banks (and perhaps other highlyregulated organizations) will have to widen bidoffer spreads to remain profitable. As a result, foreign exchange market liquidity may diminish and a substantial portion of foreign exchange turnover would migrate to less regulated entities.
COMPARISON OF MODEL PARAMETERS: THE BIS PROPOSAL AND FINANCIAL INSTITUTIONS

Financial Inst 
BIS Proposal 
Compounding Effect 

Holding Period 
1 Day 
10 Days 
3.16 
Confidence Interval 
1.65  2.0 
2.33 
1.16  1.41 
Correlations 
Across Market Factors 
Within Market Factors 
1.12 
Multiplication Factor 
1 
3 (Minimum) 
3 
Cumulative Compounding Effect 


12.1  14.7 
Chart 1 (attached) compares the capital that would be held under the latest BIS proposal against a long spot yen position of $100 million equivalent, showing profits and losses over rolling 10day periods. As the chart demonstrates, the BIS proposal would require capital of almost 15 percent ($15 million) for this position.3 The largest 10day price movement in the past ten years (June 12, 1985  June 12, 1995) was a 12 percent gain in the yen shortly after the September 1985 Plaza Agreement. The proposed BIS capital requirement is therefore more onerous than would have been necessary for this "worst case" historical experience. The second largest 10day price movement was 8 percent, equivalent to only half of the BIS proposed regulatory capital level.
It should also be noted that a yen position of this size could be liquidated in 1 day rather than 10 days. The largest 1day price change in the same 10year period was 3.47%, implying a portfolio value change of $3.47 million. The $14.88 million of capital required under the BIS proposal is 4.3 times greater than the largest historical 1day loss on this portfolio over the past 10 years.
A comparison between the BIS credit risk capital guidelines and the proposed capital requirements for market risk guidelines also leads to the conclusion that the BIS market risk proposal is unduly conservative. Committee members agree that foreign exchange trading positions (which can generally be liquidated in one day) impose less risk than longterm commercial loans (which cannot be offset until maturity). Yet the BIS market risk proposal would require capital of 14 percent or more against certain market risks while the credit risk guidelines require 8 percent against longterm commercial loans.
Multiplication Factor
The Committee understands that it may be appropriate to use a multiplication factor to transform valueatrisk figures into suitable capital levels. However, using a multiplication factor in addition to the highly conservative model assumptions in the BIS proposal generates capital requirements that are clearly excessive. As discussed more fully below, the Committee recommends that the BIS proposal include a multiplication factor but allow banks to use their own model parameters to calculate value at risk.
While the Committee is fully supportive of the qualitative standards outlined in the BIS proposals, there is concern that national supervisors, both within and between international jurisdictions, must apply a consistent approach in determining the level of each bank's compliance with the standards. This is particularly important as the results of these compliance assessments will be used to determine the multiplication factor assigned to each bank. To reduce the level of arbitrariness in this exercise, the BIS should develop a set of detailed guidelines for use by the national supervisors to ensure consistent application and measurement of compliance with the qualitative standards.
Holding Period
A 10 business day holding period is unjustifiable and ignores the fact that, even if a particular instrument is not readily marketable, its risk can often be hedged in liquid markets. Given that the large majority of both trading4 and positiontaking in foreign exchange occurs in major currency pairs, the Committee recommends that the common holding period for all currencies should be 1 day. A 1day holding period also facilitates more accurate back testing of valueatrisk calculations against actual daily revenues.
Correlation
Sophisticated institutions use correlations across risk categories (e.g., interest rates and exchange rates) to measure portfolio risks more accurately and often employ diversification strategies to reduce risks. Disallowing the possibility of any crosscategory correlations for market risk capital discourages risk reduction through diversification and reduces the reliability of potential loss forecasts.
Confidence Interval
The 99th percentile, 1tailed test (equivalent to 2.33 standard deviations in a normal distribution) is also conservative. Most financial institutions use confidence intervals ranging from 1.65 standard deviations (95th percentile) to 2 standard deviations (97.7th percentile). Combined with other highly conservative BIS proposed assumptions, a wide confidence interval can generate potential loss forecasts well in excess of actual risks.
Observation Period
The Committee strongly recommends the use of a single observation index weighted to capture the benefits of both long and short observation periods. A weighted methodology would respond to changing market environments while preserving the importance of earlier data. Committee members believe that the dual observation period under consideration by the BIS would be operationally burdensome.
De Minimus Exemption
Committee members believe that the de minimus exemption should be applied to all banks. Whether banks take positions for their own account or not is irrelevant given the exemption criteria of overall net open positions exceeding 2% of eligible capital. The de minimus exemption should also not include a requirement on the size of a bank's matched foreign exchange positions. Matched positions are already covered under BIS credit risk guidelines.
Using Internal Bank Models
The Foreign Exchange Committee strongly supports the use of internal models to calculate capital against possible losses from market price movements. However, as outlined above, the proposal as currently drafted includes minimum quantitative standards that are very different from most banks' internal models. In many instances, the proposed BIS model may be used solely for calculating regulatory capital rather than for daytoday risk management purposes.
The BIS proposal requires that banks "backtest" their past valueatrisk calculations against actual profits and losses. The Committee agrees that backtesting is a crucial element in the validation of any bank's value at risk model. Although the BIS proposal is not specific in this regard, we are assuming that the requirement is for banks to backtest their own internal models. The conservative assumptions in the BIS model would make it extremely difficult, if not impossible, for banks to backtest the BIS model. For example, to backtest a model using the proposed 10day holding period would require that the model's calculations be compared with actual revenues over a 10day period. Virtually all trading portfolios change significantly over any 10day period, making it impractical to compare the proposed valueatrisk calculations with actual revenues. In a similar fashion, using a highly conservative 99% confidence interval will make it extremely difficult to judge whether the interpreted results from backtesting are statistically significant. Back testing the BIS model would be purely a regulatory burden which would provide little, if any, benefit to the bank's risk management capabilities.
A simpler and more effective approach would allow each bank to use its own internal model to compute risk capital. Regulators could utilize the multiplication factor, if necessary, to adjust bank computed value at risk to appropriate capital levels. To evaluate the accuracy of internal models, regulators would review both the results of back testing as well as the methodologies employed in the back testing process. Internal models with poor predictive capabilities would be penalized with a multiplication factor greater than one. In this manner, regulators could encourage the development of more precise risk measurement models while maintaining consistent and conservative levels of risk capital.
Notes:
1. Confidence interval of 2.33 standard deviations * 3.16 (square root of 10day vs 1day holding period) * 1.1 (excluded correlations) * 3 (minimum proposed multiplication factor) = 24.3 standard deviations.
2. This estimate is based upon the results of a comparison conducted by a major U.S. money center bank represented on the Committee in June 1995. This bank compared its daily valueatrisk (VAR) figures using correlations across interest and exchange rate movements with modified estimates allowing no correlations. The uncorrelated estimates were consistently 1.1 times this bank's correlated VAR calculations.
3. The BIS proposed capital guideline for this position would be as follows:
 2.13% (1 standard deviation of yen change for 10day holding period) *
 2.33 standard deviations (99% confidence interval) *
 3 (Multiplication Factor) *
 $100 million (notional position) = $14,888,700
4. According to the BIS Central Bank Survey of Foreign Exchange Market Activity in April 1992, 83.2 percent of all global spot foreign exchange transactions were in major currency pairs including the US dollar, Japanese yen, Deutsche Mark, other European currencies, the Canadian dollar and the Australian dollar (Table IIb on page 10).